Run the analysis with all bandpass for power per channels

BIAPT / eeg-pain-detection

Repository for detection of pain from EEG brain signals using machine learning

5 stars 2 forks source link

Run the analysis with all bandpass for power per channels #21

Closed yacineMahdid closed 4 years ago

yacineMahdid commented 4 years ago

We know that these bandpass are important so running the analysis with all of them should be done sooner rather than later. We also need to make a module that add in automatically to the feature dataframe.

Need to calculate the power at the four bandpass:

[x] delta: 0.5-4Hz
[x] theta: 4-8Hz
[x] alpha: 8-14Hz
[x] beta: 14-32Hz

yacineMahdid commented 4 years ago

I'll run the analysis with the window = 10sec and the step size = 10sec also. This will allow me to get a glimpse of if the feature at the different bandpass are helping.

yacineMahdid commented 4 years ago

I was able to run the analysis easily with Healthy, however since we are running a lot of classifiers for the grid search this becomes very slow.

yacineMahdid commented 4 years ago

I've re-run for MSK with a model that I arbitrarily choose and we get about 64% accuracy. The real grid search would have done better than that, but it's a good temporary estimate.